Application of First-Order Markov Chain Models in Earthquake Activity Prediction
Year: 2026
Type: conference-paper
Abstract: Indonesia is situated between three different tectonic plates, making it one of the most seismically active areas. Providing exact prediction to when and where earthquake may happen remains challenging due to its unpredictability, varied factors and complex seismic processes. This study proposes a probabilistic framework that predicts earthquake activities using a first-order Markov Chain model. Historical earthquake data from Badan Meteorologi, Klimatologi, dan Geofisika (BMKG) and The United States Geological Survey (USGS) data were preprocessed and converted into discrete seismic events. States are defined based on earthquake magnitude, depth, and geographic location. The model creates a matrix that shows the probabilities of moving from one state to another to estimate future seismic activity patterns over time. Experimental Results show that the proposed approach works well in capturing short-term seismic transition behavior, including regional persistence and magnitude continuity. Top-1 and Top-3 prediction accuracies of 21.00% and 39.53% represents the improvements that are each 26.67 times and 16.74 times better than the baseline. This demonstrates that the model learns significant patterns across 127 possible states. The framework provides interpretable probabilistic forecasts with low computational efficiency, making it a lightweight baseline for analyzing earthquake sequences and combining them into data-driven approaches for disaster mitigation uses.
Institution: Binus University
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7204759882